Natural killer cells phenotypic characterization as an outcome predictor of HCV-linked HCC after curative treatments

Figures & data

Table 1. Demographic and clinical characteristics of patients and controls

	Healthy controls	Cirrhotic patients	HCC patients
#	18	12	70
Gender (M/F)	12/6	9/3	41/29
Mean age^*±SD (years)	53.61+11.65^§	56.88+11.92^§	70.57+7.92
BCLC A-0/B	na	na	61/9
HCC # 1/>1	na	na	50/20
Mean HCC size^**±SD(mm)	na	na	37.23±18.63
Child-Pugh A/B	na	12/0	60/10
CD3^+CD56^−/TL	69.78+10.22	61.27+11.32	68.98+10.52
CD3^−CD56^+/TL	12.91+6.26	14.74+9.68	10.60+7.12
Type of treatment (#)
- SR	na	na	21
- RTA	na	na	49
Median TTR^*** (mo)	na	na	16.5
Median OS (mo)	na	na	64

* age at diagnosis for patients with cirrhosis or HCC.

** sum of major diameters for each nodule

^§ significantly different from HCC patients (p = 0.0001).

*** TTR data not available for two patients without complete necrosis after treatment by RTA.

M: males; F: females; SD: standard deviation; TL: total lymphocytes; SR: surgical resection; RTA: radiofrequency thermal ablation; na: not applicable.

Table 2. TTR and OS of HCC patients according to clinical characteristics, NK-cell frequency and phenotype cluster.

	Univariate				Multivariate
	TTR		OS		TTR		OS
Variables	p	HR (95% CI)	p	HR (95% CI)	p	HR (95% CI)	p	HR (95% CI)
Median age (>72 vs. ≤72 years)	0.18	1.43	0.001	2.69	—	—	0.001	3.35
		(0.85–2.43)		(1.47–4.96)				(1.64–6.83)
Gender (F vs. M)	0.057	1.70	0.23	1.42	—	—	—	—
		(0.98–2.95)		(0.79–2.57)
Child-Pugh (B vs. A)	0.011	3.87	0.002	7.18	0.14	1.84	0.002	4.33
		(1.37–10.92)		(2.11–24.40)		(0.83–4.09)
BCLC (A-0 vs. B)	0.417	1.37	0.91	1.04	—	—	—	—
		(0.64–2.92)		(0.46–2.35)
Nodules # = 1 vs. > 1	0.20	0.67 (0.37–1.24)	0.13	0.60 (0.32–1.17)	—	—	—	—
Median nodule size (≤30 vs. >30 mm)	0.15	0.67	0.39	0.77	—	—	—	—
		(0.400–1.15)		(0.44–1.38)
Type of treatment (SR vs. RTA)	0.24	0.727	0.039	0.526			0.578	0.79
		(0.43–1.244)		(0.29–0.97)				(0.33–1.84)
CD3^+CD56^−/TL (≤ vs. > median %)	0.22	1.376 (0.82–2.30)	0.58	0.856 (0.49–1.51)			—	—
CD3^-CD56^+/TL (≤ vs. > median %)	0.73	0.9165	0.29	1.357			—	—
		(0.55–1.53)		(0.77–2.40)
NK phenotype cluster (1a vs. 1b+2)	0.031	2.01	0.0002	4.06	0.14	1.57	0.01	2.36
		(1.06–3.79)		(1.93–8.51)		(0.87–2.84)		(1.23–4.51)

Univariate analysis was carried out by the Kaplan–Meier method and compared by log-rank test. Cox proportional hazards regression model was conducted for multivariate survival analysis. Only variables significantly associated with TTR and/or OS in the univariate analysis were included in the multivariate model.

HR: hazard ratio; CI: confidence interval; F: females; M: males; BCLC: Barcelona Clinic Liver Cancer; SR: surgical resection; RTA: radiofrequency thermal ablation.

Figure 1. Phenotype and function of NK-cells in patients and controls. (A) Dot plots showing gating strategy and phenotype for CD3ζ, perforin and NKG2A in a representative healthy subject, a patient with HCV-related liver cirrhosis and a patient with HCC (left to right). (B) Differential distribution of NK-cell phenotypes and functional parameters in HCC patients and controls. CD107a expression was evaluated on unstimulated NK-cells. Significance levels are shown on top of each panel.

Figure 2. Hierarchical clustering analysis of 24 NK-cell phenotypic markers in patients with HCV-linked HCC. The heat map represents the frequency of each phenotype (higher expression in red and lower expression in green). Clustering analysis was done after median centering on both the tested markers and patients using Ward's linkage and Euclidean distance.

Figure 3. Kaplan–Meier curves of overall survival (OS) and time to recurrence (TTR) of HCC patients according to NK-cell phenotypic clustering. Left panels: comparison of patient clusters 1a, 1b and 2 (Fig. 2); right panels: comparison between patient cluster 1a and clusters 1b+2. Significance values (log-rank test) are shown on top of each panel.

Table 3. Comparison of NK-cell phenotypic subsets according to hierarchical clustering of HCC patients. Mann–Whitney test. SD: standard deviation; TL: total lymphocytes; BR: bright; NS: not significant

Phenotypes %(mean±SD)	Cluster 1a	Clusters 1b+2	p
NKCD56^+/TL	7.62 ± 5.44	11.98 ± 7.43	0.0038
NK CD56^BR	18.1 ± 1.83	7.86 ± 0.74	<0.0001
NK CD16^−	15.58 ± 6.67	7.38 ± 4.88	<0.0001
NK CD56^DIM	81.9 ± 1.83	92.14 ± 0.74	<0.0001
NK NKG2D^+	77.04 ± 12.43	81.43 ± 6.82	NS
NK NKG2A^+	74.46 ± 9.35	65.2 ± 13.37	0.0072
NK NKP46^+	74.67 ± 7.67	61.89 ± 19.06	0.0038
NK NKP30^+	69.1 ± 8.97	59.58 ± 21.33	NS
CD16^− NKG2D^+	71.75 ± 11.75	67.82 ± 14.28	NS
CD16^− NKG2A^+	90.93 ± 4.29	85.17 ± 12.2	0.025
CD16^− NKP46^+	86.31 ± 5.75	77.07 ± 16.57	0.015
CD16^− NKP30^+	63.27 ± 10.16	55.44 ± 15.87	0.04
CD16^+ NKG2D^+	78.22 ± 12.47	82.66 ± 7.01	NS
CD16^+ NKG2A^+	71.94 ± 11.13	64.25 ± 13	0.026
CD16^+ NKP46^+	73.62 ± 9.42	60.81 ± 20.14	0.005
CD16^+ NKP30^+	70.57 ± 11.47	60.43 ± 20.16	NS
NK CD3ζ^+	74.59 ± 6.48	86 ± 7.25	<0.0001
NK Perforin^+	65.66 ± 11.78	82.28 ± 7.63	<0.0001
NK Granzyme B^+	70.8 ± 9.19	80.73 ± 7.67	<0.0001
CD16^− CD3ζ^+	53.29 ± 10.7	62.36 ± 17.26	0.023
CD16^− Perforin^+	45.96 ± 13.63	53.67 ± 15.93	0.031
CD16^− Granzyme B^+	33.92 ± 13.77	33.13 ± 16.2	NS
CD16^+ CD3ζ^+	79.49 ± 8.12–	88.12 ± 7.12	<0.0001
CD16^+ Perforin^+	70.45 ± 12.67	85.4 ± 6.66	<0.0001
CD16^+ Granzyme B^+	78.86 ± 9.11	85.56 ± 6.85	0.0029

Figure 4. Co-expression of NKG2A, NKp30 and NKp46 in NK-cells from HCC patients and controls. (A) Strategy for the flow cytometric analysis of receptor co-expression in NK-cells; (B) Frequency of NK-cells co-expressing NKG2A, NKp30 and NKp46 in HCC patients' clusters and in controls. Significance levels are reported.

Figure 5. Cytotoxic function of NK-cells from HCC patients and controls. (A) Dot plots show upregulation of CD107a in NK-cells after incubation with K562 target cells with and without IL-15 activation in representative patients and controls. Top to bottom: HCC cluster 1a, HCC cluster 1b+2, healthy control, HCV-related cirrhosis. (B) Results are shown with or without overnight activation with IL-15. Percent cytotoxicity was normalized to the frequency of circulating CD3⁻CD56⁺ cells. Significance levels are reported.

Figure 6. Cytokine production in NK-cells from patients and controls. (A) Dot plots show staining of NK-cells for IFNγ and TNF-α upon IL-12/IL-18 stimulation in representative patients and controls. Top to bottom: HCC cluster 1a, HCC cluster 1b+2, healthy control, HCV-related cirrhosis. (B) Frequency of NK-cells producing IFNγ (left axis, white symbols) and TNF-α (right axis, gray symbols) was normalized to the frequency of circulating CD3⁻CD56⁺ cells. Significance levels are reported.